7.5 Million Versus 12 Million: The Hole in FBR's Filer Database the Headline Never Shows

The report from Islamabad carries a number I had to read three times — nearly 12 million people filed returns for one-time transactions, and only about 7.5 million of them are filing continuously. The gap is roughly 4.5 million. That is the number which, read as a headline, gives you a picture of 'tax evaders', and which, read through the lens of system design, gives you a database counting two different things in one column. I lean towards the second reading, because as far as I understand the Federal Board of Revenue's IRIS system has been caught for some years in a structural confusion over the 'stop-filer' category, and this story is an outward symptom of that confusion.

One thing I should clear up first — I am writing from Manchester, not from a tax office in Pakistan. I do not have the texture that a reporter in Islamabad would have — which notice went to whom, when the IRIS server slowed down, how long the queue was at which counter. What I have is numbers, and therefore every claim I make comes with its sample size and its limits attached. My one clear, falsifiable assertion is this: this 4.5 million gap is not the volume of tax evasion in Pakistan — it is a classification artefact, noise generated by counting two different populations in a single tally.

To understand this, first understand that a continuous filer and a person registered for a one-off transaction are not the same thing. In Pakistan's tax architecture, if someone bought land, bought a vehicle, or made a large financial transaction, they had to be entered in the tax system for the validity of that transaction and to file a return — whether or not they had regular taxable monthly or annual income. That is a transaction-trigger filing, not an income-based one. A continuous filer, by contrast, accounts for their income every year. Counting both in one database as if they were the same is like counting in a cricket squad the players who only played a trial match and left, and then claiming the team's bench depth is strong. From the numbers I have, it is clear that the 12 million figure includes people who made a single transaction entry and never came back. The roughly 7.5 million are those actively filing.

Now I give you the number that struck me most, and I write it the way I used to log yorker positions in my Lancashire days — according to FBR sources, these roughly 12 million one-time filers fall into the stop-filer category, and notices are being issued to them continuously. But the real discovery of the system is that the mechanism for finding out is itself hollow. Until a person reopens their IRIS account, they do not learn they are a stop-filer. This is exactly like a batsman being given out, but the scoreboard is not updated — the player standing on the field does not know he is out. Notices go by email, and email is a reliable channel to a person who maybe opens his inbox once a month. I cannot give a number here — the story does not say how many notices bounced or how many were answered. I can only say that as a mechanism, email-only notification is unsuitable for a database with a large number of inactive users.

Then comes the Finance Act 2026 provision, which speaks of restrictions on economic transactions. This is no new technique in the history of tax compliance — keeping someone out of filing a return is one thing, making it hard for them to buy land or a vehicle is quite another. The question here is who the restriction applies to. If it also applies to people with no fixed income — say a retired person who once bought a car from his savings, or a woman who filed to buy a small property in her name — then the policy will look good on a compliance metric but in practice will build a barrier that pushes economic transactions out of the formal system altogether. The provisions of the Finance Act 2026 also met political-economy constraints — stakeholder pressure, administrative feasibility, and the fear that too strict a restriction would provoke open resistance. This suggests the law was made, but its implementation is already in negotiation. I have no way of knowing whether these restrictions are fully operational or legally tested or not — and that is the limit I state plainly.

A Different Angle: The Problem Is Not the Filer, It Is the Category

What is least audible in this whole debate is that the foundation of any government machinery is its database, and if two different kinds of people sit in one category, every policy built above it rests on conjecture. If you treat roughly 12 million as filers and want to remove some of them, you will not know who to remove, because the perception will persist that all 12 million are doing something wrong. Those who think differently in this debate might advise me to suggest this — let policy first decide whether a filer is a transaction-filer or an income-filer. If you gradually split the roughly 12 million into three categories — those actively filing continuously, those registered once for a transaction, and stop-filers — then the method of sending notices improves and the number that emerges later becomes reliable too. In tax, counting is more reliable than sentiment. When you write roughly 12 million, that is a score, not a policy. Policy is made when you write how many of them actually have income.

Here I will add something from my experience. Every large system I have observed since 2026 — whether the IPL auction system or county cricket squad data — has had one thing in common: whenever an institution tries to build credibility in its numbers, the first task is to clear old and inactive entries from its database. I did not count 4,300 deliveries silently — I also noted which delivery did not exist in the universe, meaning which delivery was only in the scorebook but not in the over. The one-time transaction gap is the same. The shiny number people read first is only the surface of the policy.

Let me also make clear that none of this discussion of Pakistan's tax system connects to cricket in this article, and I am deliberately not viewing it through a cricket lens. I have learned in my career that when a story falls outside your domain, the most honest stance is to read it as if you were looking at the IRIS scoreboard, not a cricket scoreboard. Yet a parallel does exist here, and writing about it in cricket's name is not wrong. When a base price is set for a player in the IPL auction, the board does not write how many times he batted last time — it sorts him into 'spinner', 'finisher', 'opener'. In the same way, filers must be sorted. Without clear categories, the restriction above will not hold.

In 2026 I made a file of nearly 40 pages on something quite unlike a tax file, in which I claimed a player's price should be double. That claim proved right, but almost all the other prices in that file were wrong. I never published that file, but I kept this with me — that a number being right and a mechanism being right are two different things. The same test applies here. The roughly 12 million versus 7.5 million gap is a number nobody has yet settled. But that gap is not a fault in itself; it is a signal that somewhere below there is something wrong in the filing method. The question is not where the 4.5 million people went. The question is why they were never counted separately.

If I put it in one sentence: this is a question of cleaning the database, not of arrears — and until FBR stops counting the one-time transactor, I will have to ask of every number that comes, the same question I ask in my cricket column — what is the sample size, and is what has been counted what has actually been seen.